Poreplex

Added
- Added a support for basecalls from the Guppy's flip-flop model.
- `--fast5` now writes multi-read FAST5 files by default. `--symlink-fast5`
is removed as symbolic links are not compatible with multi-read FAST5s.
- Barcode accuracy is calibrated and reported for every read in
`barcode_score` within `sequencing_summary.txt` in phred-scale.
- For the dashboard view, the first part delimited by '|' of identifier in
the minimap2 index is used for display or mapping to transcript names.
- Poreplex now stops processing when it seems that the whole bunch of FAST5
files are not basecalled. (It stops on the 1000th non-basecalled read while
no read had been processed correctly.)

Changed
- The barcode demultiplexer is updated to provide calibrated accuracy predictions
and a higher robustness to irregular signals.

This release fixes the binary packages lacking the essential resource files.

Added
- Poreplex now support multi-read FAST5 files as inputs.
- FAST5 files which had base-called with the ONT guppy can be used as inputs.

Changed
- Fixed `filename` in `sequencing_summary.txt` points to wrong locations of
FAST5 files when both FAST5 output and barcoding are turned on.
- Poly(A) dwell time measurement is now more robust to pepper and salt
noises.
- Fixed BAM alignment writers which had generated broken files in
multi-threaded processing.

This minor release fixes an ABI compatibility issues in the binary packages.

Added
- Reads now can be demultiplexed without basecalling.
- With `--polya` turned on, poly(A) tail dwell time is written
to `sequencing_summary.txt`. More detailed information is
written to the dumped events by the `--dump-basecalled-events`
switch.
- Shows detailed error messages in `poreplex.log`.
- The files generated by `--dump-basecalled-events` now includes
the attributes for signal scaling parameters and the last position
that DNA adapter ends within the signal.
- Extremely short sequences are now sorted into `fail`ed reads.
The minimum required length can be changed with the
`--minimum-length` switch. (Suggested by Nathan Roach)

Changed
- Fixed a segmentation fault when using
[albacore 2.3.3](https://community.nanoporetech.com/posts/albacore-2-3-3).
- Fixed an error that stops overall process by an invalid FAST5 file.
- `filename` in `sequencing_summary.txt` is now shown as a relative path
from the output directory, not from the subdirectory for a read
group.
- Fixed a problem that separate lines of FASTA, FASTQ or
`sequencing-summary.txt` are mixed up in the output file sometimes.
- Turned off the chimeric read filter by default. Now the `--filter-chimera`
option turns it back on.
- Updated the neural network model for barcode demultiplexing for
even less false positives using randomly stitched signal fragments
as a background.